MiniMax Music 3.0
VerifiedOpen-weights music model from MiniMax that generates full songs with vocals
Product preview
MiniMax Music 3.0
What it is
MiniMax Music 3.0 is a text-to-music model from Chinese AI company MiniMax. Given a description of the style and optional lyrics, it generates a complete song in one pass, handling composition, arrangement, singing and production, for tracks up to about five minutes. Lyrics can include section tags such as verse, chorus and bridge, and a structured caption can set genre, tempo, key and vocal details. MiniMax released the weights and inference code in August 2026 on Hugging Face and GitHub under its own community license, with support in ComfyUI and Diffusers. It requires a CUDA GPU to run locally, and MiniMax points new users toward self-hosting rather than its hosted music API.
Key features
- Full songs up to about five minutes
- Section tags for verse, chorus and bridge
- Control over genre, tempo and key
- Runs locally via ComfyUI or Diffusers
What we liked
- + Open weights you can run on your own hardware
- + Generates vocals and arrangement together
- + Fine control through structured captions
Worth noting
- · Needs a capable CUDA GPU and technical setup
- · Custom community license to check before commercial use
Editor's notes
More context from AI BirdView
Most capable song generators live behind a subscription and a web interface. MiniMax Music 3.0 stands out because MiniMax published the model weights, so developers and hobbyists with the right hardware can generate full songs with vocals on their own machines. It aims at complete, structured tracks rather than short loops, which makes it interesting for anyone exploring AI-assisted songwriting.
- Produces complete songs of up to about five minutes in a single generation
- Lyrics accept section tags, so verses, choruses and bridges land where you want them
- Structured captions can specify genre, tempo, key, vocal style and arrangement
- Weights and inference code are published on Hugging Face and GitHub
- Works with ComfyUI and Diffusers, and MiniMax documents a self-hosting path
MiniMax Music 3.0 is best for developers, researchers and technically comfortable musicians who want to experiment with AI music locally or build it into their own tools. MiniMax says the hosted music API is closed to new users, so self-hosting is the main route. A practical tip: read the community license before using output commercially, and start with short lyrics and a detailed style caption to learn how the model responds.
Reader reviews
From people who've used MiniMax Music 3.0. Emails are never shown publicly.
No reviews yet. Be the first to share your experience.
Alternatives to MiniMax Music 3.0
Other tools we like in this space.
Suno
Radio-ready songs from a text prompt
Udio
Studio-grade AI music with fine-grained control
AIVA
An AI composer for cinematic scores
Stable Audio
Stability AI's music and sound generation studio